This talk introduces an application of data assimilation (DA)---a nonlinear regression that integrates knowledge by regressing using a nonlinear mechanistic physiologic model---to forecast glucose, insulin and infer both externally validated and unmeasureable physiology in the ICU setting. The DA estimated glucose well for some individuals and inferred parameters, e.g., kidney and liver function, were consistent with laboratory measurements, demonstrating the potential of DA as a tool in the ICU and for understanding pathophysiology.
Learning Objective 1: The first learning objective is to understand what data assimilation is and how it can be used in a clinical and basic-science setting using clinically collected data.
Learning Objective 2: The second learning objective is to understand how, in our preliminary work, data assimilation worked when used to infer glucose and other physiologic functions in an ICU setting.
David Albers (Presenter)
Matthew Levine, Columbia University
Andrew Stuart, Califonia Institute of Technology
Jan Claassen, Columbia University
Bruce Gluckman, Pennsylvania State University
George Hripcsak, Columbia University